BiasMix-Finance：面向LLM投资组合建议的生成后KYC护栏
BiasMix-Finance: Post-Generation KYC Guardrails for LLM Portfolio Advice
- Vizuara（维祖拉公司）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本文提出BiasMix-Finance生成后KYC护栏流水线，含JSON模式约束、数值上限验证及凸二次规划投影，可将LLM投资组合建议违规率降至0%，并发布相关基准与代码。
AI中文摘要：
大型语言模型（LLM）可生成看似合理的ETF投资组合，却在静默中违反风险、费用和多元化方面的基本KYC（了解你的客户）式约束，这在多轮智能体咨询系统中尤其成问题，因为每一条建议草案都可能成为实际行动，除非有可审计的执行层加以防护。本文研究一种与模型、资产无关的生成后护栏流水线：（i）执行严格的JSON分配模式；（ii）验证分配是否符合数值上限；（iii）当出现违规时，通过凸二次规划（QCQP）确定性地将输出投影到最接近的可行投资组合。本文推出BiasMix-Finance（Mini），这是一个紧凑的压力测试基准，用于应对LLM生成存在偏差时的约束决策，包含16只ETF的 universe、三类投资者画像及八个偏差提示。在三种模型和三种推理模式（直接、批判、自洽）下，首轮生成在47.6%-85.7%的测试用例中违反至少一项上限（合并值为67.2%），但凸投影层将最终可行性违规率降至0%，且仅需极小的修正距离（测试合并中位数D=||w*-w0||_2=0.066），表明护栏通常能保留原始分配的意图。本文报告了带置信区间的违规率与修正距离，以及经多重检验校正的配对模型比较。为支持可复现性，本文在公共GitHub仓库发布了数据集、提示、上限及代码。
英文摘要：
Large language models (LLMs) can generate plausible-sounding ETF portfolios while silently violating basic KYC-style constraints on risk, fees, and diversification. This is especially problematic in agentic multi-turn advisory systems, where each draft recommendation can become an action unless guarded by an auditable enforcement layer. We study a model-agnostic, asset-agnostic post-generation guardrail pipeline: (i) enforce a strict JSON allocation schema, (ii) validate allocations against numeric caps, and (iii) when violations occur, deterministically project the output to the nearest feasible portfolio via a convex quadratic program (QCQP). We introduce BiasMix-Finance (Mini), a compact stress-test benchmark for constrained decision-making under biased LLM generations, with a 16-ETF universe, three investor profiles, and eight bias prompts. Across three models and three inference modes (direct, critique, self-consistency), first-pass generations violate at least one cap in 47.6-85.7% of test cases (67.2% pooled), but the convex projection layer reduces final feasibility violations to 0% while requiring only a small correction distance (test pooled median D=||w*-w0||_2=0.066), indicating that the guardrail typically preserves the intent of the original allocation. We report violation rates and correction distances with confidence intervals, and paired model comparisons with multiple-testing correction. To support reproducibility, we release the dataset, prompts, caps, and code in our public GitHub repository.
